Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://xmm.vilspa.esa.es/calibration/documentation/epic_cal_meetings/201504/Lumb_Cal_2015.pdf Äàòà èçìåíåíèÿ: Wed Apr 8 18:31:52 2015 Äàòà èíäåêñèðîâàíèÿ: Wed Apr 13 13:28:09 2016 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: ï ï ï ï ï ï ï

Cross-calibration investigations ­ straylight (and others)
D Lumb


Motivation
· Following the last EPIC calibration meeting continued investigations into off-axis stray light · Aim is to understand reported gross discrepancies, and examine if a physical model also contributes to understanding on-axis cross calibration uncertainties


Method
· Looked into archive for observations with obvious single reflection arcs · Dominated by LMCX-1, Cyg X-3 and Galactic Bulge binaries · As the arcs are very extended across the focal plane have reanalysed all using the ESAS scripts · Made a comparison with SCISIM ray trace that replicates each observation


Main Analysis Steps
· Select regions for each camera and run esas · Extract spectra and background · Made an .arf that is a "guess" for typical off axis angle ( fn(E) knowing only a single reflection occurs ) · Assume source spectrum of a LMXRB e.g. const*phabs*(diskbb+comptt) probably the detailed spectrum doesn't matter to 1st order · Also require an in field background of gauss+gauss+apec+phabs*apec that has a different normalisation per camera


· Depending on epoch - find nearest RXTE ASM or MAXI observation (2-10keV flux or sometimes for latter even a spectrum ) · With MAXI data joint spectral fit with a constant multiplier between instruments · Or do EPIC spectral fit alone and just make a flux estimate comparison · Difference gives the straylight rejection factor · If there is a gross problem in any source, should at least be able to renormalise the 3 cameras to allow azimuthal trends to be tracked

Source flux


Ray Tracing Tasks
· Assume catalogued source position and the RA,DEC, PA PNT keywords to define a source input to SCISIM · Ray trace for all 3 telescopes · Especially for smaller off-axis angles, realised need a good estimate of focal plane offsets and geometries to determine fraction of single reflected arcs contained within observation selected regions (not in plain vanilla SCISIM) · Now realise that to get enough sources will need to use full range off-axis radii, where the "guess ARF" is NOT good enough ( more work )


Factors also considered
· By far biggest uncertainty is the source flux (RXTE 5% systematic on its Crab calibration, but sometimes are no simultaneous data) · Statistical errors on flux ~5-30% typically · Unknown effect of spectral model and the use of assumed .arf ­ but probably a common factor · Compared average of observation with average of simulation to flag gross source flux problem · ~50 observations continued for final analysis


Radial Variation
· Nominal radial variation from SCISIM (circular focal plane FOV)


Radial Variation
· Radial variation for the pathology of observations ­i.e. when you add in the focal plane geometry


Azimuthal Variation
· Ray trace for a "typical" 0.8 degree off-axis source and uniform azimuthal distribution · Compared for both a 100 micron baffle shift or a 0.02° Rotation, onto a circular aperture in detector plane · Typical change +/- ~cosine like variation


Camera results - radial · Average of 3 cameras


Scaling for source flux error?
Average 3 cameras difference between measured and simulated


PN Camera results - azimuthal


PN Camera results ­ ratio to simulation


PN Camera results - azimuthal
Scale for flux errors and divide by simulation

?


MOS1 Camera results - azimuthal


MOS1 Camera results ­ ratio to simulation


MOS1 Camera results - azimuthal
Scale for flux errors and divide by simulation

?


MOS2 Camera results - azimuthal


MOS2 Camera results ­ ratio to simulation


MOS2 Camera results - azimuthal
Scale for flux errors and divide by simulation

?


Conclusions ­ still in progress
· Average straylight vs radius not grossly discrepant · After source flux adjustment, different azimuthal variation per camera · Not possible to say if due to baffle rotation or shift · Possible impact on Effective Area on axis (Fn of Energy) was ray traced ­ only ~1% effect · Azimuthal variation ­ like the Matteos effect · But also has similar effect to the telescope axis modification made XMM-CCF-REL-156 in 2004 ? · Is that a combination of two effects??????


Questions about ARF
· Work on clusters cross-calibrations and the XMM Catalogue point source comparisons show the similar effects · XMM gives lower temperature than ACIS, PN effective area returns progressively lower flux than MOS · Few % effect ­ MG wondered if in the end we just use an aphysical fudge factor


ARF vs extraction radius
· Using SAS 13.0 ( as likely was done for pre 2014 IACHEC cross calibration exercises) · Generate an ARF for different spectral extraction radii up to ~10 arcmin · As expected compares quite well with the Encircled Energy output of calview when using ELLBETA accuracy · But seems very aphysical ­ why doesn't mirror scattering get worse with increasing energy?


PN ARF

Mirror scattering should decrease ARF with energy


MOS ARF

Less obvious increase (except E>8keV)


Encircled energy differences
· 0.5 ­ 2 keV band and "typical" enclosed diameters 54 arcsec & 1.67 arcmin
PN 86.3 / 93.5% MOS1 83.0 / 90.9%

· 2 ­ 7 keV band and "typical" enclosed diameter 54 arcsec & 1.67 arcmin
PN 87.1 / 93.8% MOS1 82.8 / 90.6%

· PN would report 3- 5% lower flux yet actually its PSF is worse than MOS1 (admittedly core)


PN Encircled Energy Normalised to ~10 arcmin radius


PN ARF matches the ELLBETA description

Small (<1%) differences easily explained by chip gaps, bad pixels etc. ?


OOT Correction
· Seems to do nothing for the withootcorrection flag in arfgen ? · Exposure is supposed to be corrected for ~6% OOT effect, but comparison with GTIs and LIVETIMEs shows much larger reduction? · In any case with extended sources how do you bookkeep for increasing recovery of OOT events INTO spectral extraction? · PN has more complex time accounting ­ any small errors potentially affect normalisation · And for clusters the cross-arf problem will surely blow up any incomplete ARF accounting as fn(E) !!


Example:
PN MOS1 76317

Observation Duration

69719

On Time Live Time Total GTI Image Exposure
Spectrum Exposure

64122
56489

68914
68012

64124
56359

68921
68002

Spectrum/Image Exposure

0.88

0.987


Flux and Spectral Differences?
· We are looking for few % differences in normalisation and also some trend with energy · Normalisation could be hostage to small errors in the LIVETIME/GTI calculations and OOT correction · Encircled energy function calculated by ELLBETA seems to go in wrong direction in energy AND between PN and MOS